Search Results for "JVET-M0447"

Found 3 document(s)

Search documents

Use number, keyword, author, or MPEG number. Filter by meeting when needed.

13th Meeting: Marrakech, January 2019 2019-01-03 02:36
Abstract
Decoder side motion vector refinement (DMVR) techniques were proposed to refine the motion vectors at decoder side after parsing stage. As intra prediction may depend on the results of its spatial neighbor’s full reconstruction, it cannot be started until the MV derivation process and the final reconstruction are finished for the neighboring blocks, which leads to serious latency in hardware pipeline. It is reported that the issue can be resolved with the method proposed in this contribution.
JVET-M0447 CE9: Constrained intra prediction with DMVR (test 9.2.4) [M. Xu, X. Li, S. Liu (Tencent)]

This contribution presents DMVR simplifications based on VTM3.0. Firstly, only 4 integer precision surround check for certain condition, followed by MRSAD mean value calculation by sampling process, then use SAD calculation to replace MRSAD. The proposed technologies have 0.71%/ 0.68%, 0.60% gain respectively compared to VTM3.0 anchor.

Decisions
This contribution presents DMVR simplifications based on VTM3.0. Firstly, only 4 integer precision surround check for certain condition, followed by MRSAD mean value calculation by sampling process, then use SAD calculation to replace MRSAD. The proposed technologies have 0.71%/ 0.68%, 0.60% gain respectively compared to VTM3.0 anchor.
Citation
13th Meeting: Marrakech, January 2019 2019-02-11 08:21
Abstract
The purpose of this Core Experiment (CE) is to investigate the coding tools related to neural network (NN) based loop filter on top of the Test Model. In this CE, the impact of the position of the NN based filter in the filter chain, the benefit of the CTU level NN filter adaptive on/off switching, and the generalization capability when QP is different from the training QP will be investigated. In addition to the encoder and decoder run-times, the network information, e.g., network parameter numbers, number of layers, and convolution kernel size will also be evaluated. The list of tools and experiments to be performed are detailed in this document.
JVET-M1033 Description of Core Experiment 13 (CE13): Neural-network based loop filtering [Y. Li, S. Liu, K. Kawamura]

Future meeting plans, expressions of thanks, and closing of the meeting

Future meeting plans were established according to the following guidelines:

  • Meeting under ITU-T SG 16 auspices when it meets (starting meetings on the Tuesday of the first week and closing it on the Wednesday of the second week of the SG 16 meeting – a total of 9 meeting days), and
  • Otherwise meeting under ISO/IEC JTC 1/SC 29/WG 11 auspices when it meets (starting meetings on the Wednesday prior to such meetings and closing it at lunchtime on the last day of the WG 11 meeting – a total of 9.5 meeting days).

In cases where an exceptionally high workload is expected for a meeting, an earlier starting date may be defined.

Some specific future meeting plans (to be confirmed) were established as follows:

  • Tue. 19 – Wed. 27 March 2019, 14th meeting under ITU-T auspices in Geneva, CH.
  • Wed. 3 – Fri. 12 July 2019, 15th meeting under WG 11 auspices in Gothenburg, SE.
  • Tue. 1 – Wed. 9 October 2019, 16th meeting under ITU-T auspices in Geneva, CH.
  • Wed. 8 – Fri. 17 January 2020, 17th meeting under WG 11 auspices in Brussels, BE.

The agreed document deadline for the 14th JVET meeting was planned to be Tuesday 12 March 2019. Plans for scheduling of agenda items within that meeting remained TBA.

WG 11, the local hosting organization Ecole Mohammadia d’Ingénieurs (EMI) – Mohammed V University, Rabat, Morocco, and Abdellatif Benjelloun Touimi were thanked for the excellent hosting and organization of the 13th ...

Decisions
Jian Qing Zhu (Fujitsu R&D Center)
Citation
13th Meeting: Marrakech, January 2019 2019-01-10 17:03
Abstract
The tools in the scope of this CE include bi-directional optical flow (BDOF) and decoder side motion vector refinement (DMVR).
JVET-M0029 CE9: Summary report on decoder side motion vector derivation [X. Xiu, S. Esenlik]

The core experiment summary report is organized into 2 sub-tests as follows:

  • CE9.1: BDOF design (3 tests)
  • CE9.2: DMVR design (24 tests)

CE9.1: BDOF design

#

Description

Tester

Cross-checker

9.1.1.a

1. use 8-tap DCTIF filters to generate the prediction samples in extended region and inside CU

2. padding for reference samples outside [w+7, h+7] for final MC (DCTIF)

Xiu, Xiaoyu

H. Liu

9.1.1.b

1. use integer positions to generate the prediction samples in extended region

2. use 8-tap DCTIF filters to generate the prediction samples inside CU

Xiu, Xiaoyu

H. Liu

9.1.1.c

1. apply different gradient calculation method for prediction samples on the CU boundaries and inside CU

2. use 8-tap DCTIF filters to generate the prediction samples inside CU

Xiu, Xiaoyu

H. Liu

VTM

Cross-Check

Test

Document

Crosschecker

Y

U

V

EncT

DecT

EncT

DecT

9.1.1 a

JVET-M0487

H. Liu

-0.01%

0.03%

0.02%

103%

102%

105%

104%

9.1.1 b

H. Liu

0.05%

0.02%

0.02%

99%

98%

99%

100%

9.1.1 c

H. Liu

0.10%

0.03%

0.01%

99%

98%

100%

100%

9.1.1.a does not simplify, replaces bilinear filters in extended region by DCTIF

9.1.1.b simplifies by using no interpolation in extended region (just integer positions)

9.1.1.c changes the gradient calculation at boundaries and does not need extended region any more, it simplifies, but the design becomes less unified

Several experts supported 9.1.1.b as the best simplified design approach.

Decision: Adopt JVET-M0487 (solution 9.1.1.b) which uses integer positions to generate the prediction samples in extended region, and uses 8-tap DCTIF filters to...

Decisions
adopted
Adopt JVET-M0147 with SAD cost function, and without the MVD based early termination check
Citation
New Search